Detection and Recognition of Characters in Place Name Board for Driving

The objective of this paper is to detect and recognize the characters in the signage’s (traffic place name board). It is applicable especially for Indian conditions. It detect the green colored signage’s from the background using seamearing algorithm and extract the detected signage’s board then convert it into binary image. Next, it detect the signage characters from signage’s using horizontal segmentation and vertical segmentation algorithms, it extract each individual character from the signage’s. Then, the features are extracted from individual character of signage’s using DCT, DWT and Hybrid DWT-DCT. In training phase, 324 discrete wavelet features are extracted from 36 characters(9 features extracted from 36 character) in DWT, 20 highest energy coefficients are extracted by using DCT and 20 highest energy coefficients are extracted using Hybrid DWT-DCT. Finally, the extracted features from each characters are recognized using SVM. Selection of feature is probably important factor to achieve high performance in recognition. The application of this paper is a driver assistant system, to guide the driver while driving , traffic safety by calling the driver’s attention to the presence of key traffic information board. The performance of signage recognition is evaluated for place board image and the system achieves a recognition rate of 94.44% using DWT , 91.66 % using DCT, 97.22% using Hybrid DWT-DCT and SVM.